Welcome to the Software Renaissance. Your Strategy Isn't Ready | Martin Eriksson

Productized Podcast: The next person and the next talk. It's ⁓ an exciting one, it's the last one. So I ⁓ I'm starting to feel the blues, you know, that it's sweet and sour moment where we announce the last speaker of the event of the year. although we have a surprise at the end, so bear with us.

And we also have the QA. So by the way, if you haven't posted a question to Rich, you can use the code even during Martin's talk and you can also post questions to Martin. Okay, so be that this is going to be the last QA that we have available. So enjoy it, take use it.

Okay. So for the last moment that we with the speaker, we have in fortune to get and how many of you do have a copy of the decision stack? Guys. It's mostly the front people that got the books.

I I don't know what don't know what's happening, but are you guys a crew and and a gang or a mafia? I don't I don't know. Something is going on. What I would like to ask you, because this is going to be the last talk, and then we will ask to take a photo of everyone as as we end it up.

If the folks and that is generous enough, if we could left up and just populate here the spaces that we are missing In the second row, in the third row, in the fifth row, you have a lot of empty spaces. So the ones that are above that I almost cannot see you, if you guys could come below to these rows that are a bit more empty, just fill them in, please. Okay? Don't be shy, come in.

Yeah, yeah, yeah, yeah. So if you can come run up. Okay? As Matt shared, let's bring in the human communication closer together.

Okay, don't be afraid to talk with other people. Very cool. Okay. So we are getting ready for the last one.

this last person is actually or people like this last person. Other reason why I also like my kids to travel, and really a advocate that people should go outside of the countries to visit, to experience new cultures. To get that feeling that the world is not as small as people think. And I do believe that this is one of the reasons why this person has achieved so much.

So he was born in Sweden, and then he crossed from Kenya to Iraq, to Thailand, to Turkey, and then finally found the homelands in London. Okay? So with all of this travel and with a thinker curious mind, I do think that this is one of the reasons that this person was so. So successful.

sorry, my notes are crapping. Nice. Okay, and I do have a feeling that I know this person, although we only met yesterday, because my LinkedIn feed has been flooded with Martin and his new book for the last, I don't know, five, six weeks. I had the feeling that I've been stalked with ⁓ everyone that is just buying the book, sharing the book.

So he's doing an awesome job online. to make sure that people understand this. So without further ado, the last speaker of today, enjoy it. I'll be back after just to annoy you a little bit more with QA.

Okay. The author of the decision stack, Martin Erickson. Thank you very much for that introduction and welcome. we've heard a lot about AI today.

Even the people not mentioning AI mentioned AI. And before you came here, you were probably doom scrolling through your feeds and seeing lots and lots of horrible news. Because the AI panic is real. Earlier this year, a macro research shop called Citrini, something nobody had ever heard of before.

Wrote a blog post about a hypothetical future recession that was driven by AI. And it wiped hundreds of billions of dollars off the stock market. A single viral post on LinkedIn, on LinkedIn people. was seen 80 million times and wiped another couple of hundred billion off SaaS valuations.

This is the hype machine working. But the evidence isn't really there yet. This is ignoring the facts. We've seen the same panic in every technology disruption in history.

The predictions of imminent job are almost always wrong on timing and wrong on direction. And in fact we're seeing hiring for software engineers going up. It turns out that we are terrible at predicting the future. In twenty sixteen, Jeff Hinton, arguably one of the most important figures in AI, a man who would later go on to win the Nobel Prize for his contributions to AI, stood in front of an audience and made a prediction.

I think if you work as a radiologist, you're like the coyote that's already over the edge of the cliff, but hasn't yet looked down so doesn't realise there's no ground underneath it. people should stop training radiologists now. It's just completely obvious that within five years deep learning is going to do better than radiologists because it's going to be able to get a lot more experience. it might be ten years, but we've got plenty of radiologists already.

Well, that sounds terrifying. I don't want to be a radiologist running out over that cliff edge. again, this is not some random pundit. This is not a LinkedIn post.

This is not a random blog post. This is the godfather of deep learning. And he wasn't wrong about the technology. AI is really, really good at reading medical images.

But what happened to all those radiologists? It's been 10 years. Well, turns out they're actually doing better than ever. In the US alone, are 16% more radiologists today than in 10 years ago.

And you might think, well, ⁓ they're getting all this work done, maybe the price has gone down. But no, they're actually earning 44% more than they did 10 years ago. So what happened? Well again, AI has turned out to be ridiculously good at diagnostic imaging.

It's at least 50% faster than humans, it's at least 10x better than humans. And that has driven down the cost of testing. But because the cost has gone down, demand has shot up. And we have the largest radiologist shortage in history.

It turns out that when we do something cheaper, faster, and more efficient, we don't just do the same amount, we do vastly more of it. And we've seen this before as well. In 1865, an economist called William Stanley Jevons noticed something strange. James Watts' new steam engine was dramatically more efficient.

But instead of coal consumption going down, shot up. Because you make something cheaper, faster, more efficient, you don't do the same amount for less money, you do more of it. This is called the Jevons paradox, it's playing out everywhere AI touches. And we have to remember that AI is only getting started.

OpenAI may have been founded 10 years ago, Anthropic six years ago, but the moment that really launched this, that ⁓ the Gen AI craze, was just three years ago. And today's AI is the worst that you're ever going to work with. For some reason we're still typing. We're basically in the DOS prompt era of AI.

So we have to imagine everything that came after this. Windows, imagine Mac, imagine mobile, imagine apps, imagine wearables, wireless, Wi Fi, all the things that came after this moment are still to come in AI. And it's only getting faster. The speed of adoption is so ridiculous we can barely measure it.

You can see this tiny little orange chart over here shows that Gen AI has reached about the same user penetration or adoption. As the internet did in two years instead of twenty. And the cost is going down as well. Sam Altman said the cost of using any given level of AI falls about ten X every twelve months.

This is true for every model. Might not feel that way when you look at your anthropic bill every month. But again, for any given level of AI. if you're using Opus, that price is lower than it was last year.

It's just the latest stuff that isn't. We've seen this before, too. Google made the same bet in 2004 when they launched Gmail. When they launched, Gmail came with one gigabyte of free storage.

Sounds like nothing today, but at the time were offering two or four. Megabytes. People literally thought this was an April Fool's joke when they launched. But Google wasn't betting on the storage cost of the day.

They were betting on the storage cost of the future. When they started planning this in 2000, storage cost about $12 a gigabyte. By the time they launched, it was closer to a dollar. And today it's less than one cent.

They bet on the cost curve and the cost curve one. So we have to ask ourselves the same about AI. What if models become commodities? What if the cost of a token goes to zero?

What if the cost of software goes to zero? These aren't hypotheticals. This is the trajectory that we're on. There are assumptions in here around energy costs, there's localized costs that are problematic, but this is the trajectory.

And we already know what happens when we make something cheaper. We unlock demand that was always there but couldn't be served. Because it turns out that demand for solving human problems is essentially infinite. So we have to start thinking of AI as an opportunity.

This is not about replacement, this is not about less demand, this is not about doom. It's about more demand, opportunities, new frontiers. I believe a software renaissance. Wait, you I hear.

Isn't software dead? Twitter says SAS tools are dead. Okay, I like that. We could bycode our own tech stack.

I don't want to. I built a scheduling tool. We have a scheduling tool. Now it's now it's not working.

payments. No. ⁓ are not gonna pay two point nine percent and thirty cents on every transaction. No Hey, you just shared all our users' credit card information.

⁓ you're still here. Get out, you're fired. Sounds good. How much your API costs?

Mm, about five hundred? How much was the tech stack? About two hundred. Hmm.

Hey, you. Build a new CRM. So I'm sure some of you heard those conversations. I I've heard conversations not too far from that.

the build-by shift is real, right? That calculation has changed, but it's smaller than you think. Because it's only really tech companies that are thinking this way. And we all live in a tech bubble.

Every square here is about eight is about three million people. And so out of the eight point one something billion people on the planet, eighty-four percent have never even used Gen AI. And only not point not four percent. Have tried coding tools.

Most of the world simply doesn't want to build software. They want to do the thing that software enables. As April Dunford said, customers don't buy AI, they buy the elimination of a pain point. And even Anthropic is using Slack, Workday, Figma, and more.

And they've publicly said that they're gonna keep doing that. Even though they have access to the best models, they have access to Thousands of amazing engineers, they could build this, but they don't want to. And neither was a restaurant manager, an accountant, estate agent, logistics coordinators, all the other millions and billions of people out there who use software every single day. So what does this mean for us?

Well, I've been doing this for so long you can tell by my gray hairs, and I've advised over a hundred and fifty portfolio companies over the last sort of ten years. what I've learned in that time is that the only constant is change. Except that was Heraclitus two thousand five hundred years ago, not me. In most organizations, we have some sense of where we're going.

We have a vision, we have a purpose, we want to figure out how we're gonna get there. And today, sadly, the answer is AI everywhere. But even worse, we're looking at a piece of work in front of us and we're like, why are we building this AI thing? Well, we have to token max.

Why are we doing that? I don't know. So I believe every organisation, every product team to be able to answer five simple questions. Where are we going?

How are we going to get there? What matters right now on that journey? What actions do we take to move forward? And how do we choose between those actions?

I call this the decision stack, and I use vision, strategy, objectives, opportunities, and principles to answer those questions. Now, this stack of decisions is incredibly important. Because from the top down, simply answering the question how? Here's our vision.

How are we gonna achieve that? Here's our strategy, and so on. But from the bottom up, for every individual contributor, we can ask why? Why are we doing this?

Because we're trying to achieve these objectives. Why is that important? You want to achieve the strategy, and so on. It's also important because it is a stack of decisions.

As you are making decisions about your vision, you are just as important what you're saying no to as what you're saying yes to, and so on all the way down the stack. And it helps highlight when things don't make sense, if we have an opportunity in front of us that doesn't connect to something. When we have an objective that doesn't connect to our strategy, or even more importantly, when something's completely missing. Every organization has a decision stack, whether you call it that or not, but too many look a little bit like this.

It's a vision that's in a PowerPoint slide somewhere about three years ago. There's a strategy in another deck on a shelf somewhere else. There's in a spreadsheet somewhere else that we haven't looked at since the last OKR update. And then someone slots in an AI strategy.

And none of it connects. And that's what I'm seeing. Too many companies are applying AI only at this opportunities layer. They're adding a new feature here, an AI integration here.

We're sprinkling AI in all the wrong places. And then it gets to a board level and we start having a conversation about how do we already do but cheaper? That's using once-in-a-generation technology to shrink yourself. If you had woken up 10 years ago and your engineering team was suddenly 10 times more productive, would you have fired 90% of them?

Hell no. You would have been doing 10 times more. And my favorite story about this is IKEA. When they started rolling out their AI customer service bot, they saw a dramatic reduction in the need for their customer service team.

Now, any other corporate would have been like, great, we can fire 55% of our team. Let's do that. Let's save the cost. But IKEA started decided to think differently.

They realized that the other 50% of that work was really about design. And so they upscaled their entire help desk and Train them to be interior designers. And within its first year, that new business line generated one point four billion dollars of new revenue. Same people, same humans, new work.

So AI is the biggest growth lever your business has ever had, and if you treat it as c a cost lever, you're simply practicing malpractice. We have to stop treating AI as a tool, we have to stop treating it as efficiency, and we have to start treating it as a multiplier. And that means moving up the stack and thinking much more strategically. As Rita McGrath said, in a fast-moving world, strategy is more critical than ever.

And there's no better story of this than Intercom. If you don't know Intercom, they're one of the European SaaS darlings. It's a help desk software tool in much the mold of all other SaaS software tools. They grew amazingly well.

Famous for their product culture, famous for their product itself, they were doing incredibly well. Until growth started flattening just before the pandemic, the old CEO and founder came back into the business in 21. then in 22, ChatGPT landed. And they realized that not everyone was gonna be as scrupulous as ZKI.

Not everyone was going to reskill their help desk team. And Intercom's business was in trouble. Because they saw help desk teams. Software.

They sell it by the seat. And suddenly, if you're gonna fire 50% of your help desk, that's 50% of intercom's revenue. So they could have done what everyone else did, bolt the chop bot on, keep for seat pricing, and hope for the best. But instead, they reinvented.

They built a completely new brand and product called Fin, a brand new from the ground up chatbot that directly resolves customer queries. It's resolving over a million customer tickets a week. But the real move wasn't the technology. It was the business model.

So we heard from Emmanuel earlier today, they flipped it from perceiv pricing to an outcome based r ninety nine cents per resolution price. So customers only pay when the AI actually solves the customer problem. So think about what that means strategically. Their old model meant more customer problems equals more seats, misaligned incentives.

Their new model is better resolution equals more revenue. And revenue started growing again. Fin alone went to over 100 million ARR in less than a year. And then they started adding on other products.

They have an e-commerce chat bot, they have a sales pipeline chat. All adding on, building on that expertise. And then a few weeks ago they rebranded and actually just adopted the name Fin. And earlier this week they were acquired by Salesforce for nearly four billion dollars.

Another example I love is a small company called Kive from Sweden that we invested in when I worked at EQT a few years ago. The founder is a formal film director, and he used to work on these really big shoots for brands. So imagine you go out to a location, you bring the product with you, you bring a whole team with you, you're doing a bunch of film shoots, you're doing a bunch of photo shoots, and you come back with millions of assets, and you have this massive asset management problem.

And so Kive was founded just to solve that. They were using machine learning, remember AI, before AI was cool, to solve that problem. And we invested in them and they were growing well, they were building amazing stuff. But then when ChatGPT landed, the founder took a step back and had a rethink.

Because he realized that most brands were just going to skip those steps and go straight from planning to delivering a generated asset. And he wanted to be where that happened, so he completely rebuilt Kai from the ground up as a brand safe, copyright safe. brand generation tool. And so today Kai serves companies like Polestar, where they instead of having to fly a car from the factory in China to South Africa, do a photo shoot, they can just generate the assets they need from their desks in Soccer.

So the landscape has changed. As we've learned from Jevons, when something is cheap and easy to make, we want more of it. But when something is cheap and easy to make, everyone is going to make it. All the big SaaS companies are laying off great people.

They're gonna go build new companies. But also every niche that wasn't profitable to go after is suddenly competitive. So I think we're about to see an explosion of software. Competitive advantage is going to be more transient than ever.

And that means that the moats have changed. If you haven't read The Seven Powers by Hamilton Hellmer, I highly recommend it. It's a fantastic book about how to build those moats. But they're definitely changing.

Scale economies, processing power, and switching costs are more or less moot because of AI. Instead, we have to think about distribution, marketing, go to market. Network economies, existing users, capture markets are incredibly valuable at this point. But so are cornered resources like proprietary data, domain expertise, deep customer understanding, and most importantly, I think trust and brand.

But none of these have anything to do with technology. None of these have anything to do with AI. All of them are strategic investments that compound over time. So we have to think strategically and we have to think bigger.

We have to stop asking how do we add AI to what we already do? We have to stop asking how do we use AI to cut costs? And instead start asking how can we solve this customer problem in a completely new way. Not just incrementally better, fundamentally different.

The way AI powered imaging didn't just make radiology faster, it made preventative screening possible at scale. We have to start asking can our business model change completely? We got a master class on that from Emmanuel earlier, right? Like intercom, can we move from a per seat to a resolution based outcome pricing?

We have to start asking what demand is currently locked up that AI could unlock. This is the Jevons paradox question. What would your customers do with your product if it was ten times cheaper, ten times faster, or ten times more accessible? Fundamentally, we have to start asking what we can now build that was impossible before?

And then we have to bring our organization with us. AI is making everything faster, but I like to argue that speed was never our problem. As Peter Drucker said, there is nothing quite so useless as doing with great efficiency that which should not be done at all. The speed that AI gives us is a bit like setting jet fuel on fire on a runway.

It's really spectacular. You can warm your hands for a bit, but it's not really helping you get anywhere. Velocity is speed with direction. That's when you're taking that jet fuel, you're putting it in a plane, and you're starting to go somewhere.

And momentum is velocity times mass. This is when you bring your organization with you on that journey. So you have to build velocity through strategic direction. You have to build momentum through alignment.

And only then can you move up the stack and tackle becoming AI disruption. But remember in that that a computer can never be held accountable, so it should never make a management decision, especially about Voltron. So what does the future hold? So, if I got this right, smart enough to invent AI, dumb to need it, and so we can't figure out if we did the right thing.

We don't know. Let's be honest with ourselves, there's going to be change, there's going to be disruption. I've been laid off multiple times in my career, and it's never been fun. I'm not going to stand up here and pretend otherwise.

But I don't think it's all doom and gloom. I think there's a massive gap between one founder and an army of agents, as some of the really red pilled founders are talking about, and needing 10,000 employees to run a mid sized SaaS company. The reality for us will land somewhere in between. The question isn't whether disruption's coming, it's whether we take it as an opportunity to actually build something better.

Because every major technology shift has followed the same pattern. It starts with predictions of doom, whether it's horses when cars came along, bank tellers when ATMs came along, or radiologists when AI came along. It's followed by short-term disruption, which is real, it is painful, it is uneven. But it's also always followed by long term abundance, demand, new roles, bigger opportunities.

We're somewhere between one and two right now, and the temptation is to stay focused on that disruption, to panic about what AI might be taking away, to doom scroll our way through our days. But remember, we're terrible at predicting the future. A Nobel Prize winning computer scientist got it wrong, so I may get it wrong as well. Which means you have to plan for multiple futures.

Strategy is ultimately about scenario planning. If your strategy only assumes one version of the future, it's fragile. The companies that survive the coming decade will be the ones that build for multiple futures at the same time. So you have to make scenario planning a strategic discipline, not a once in a decade off site exercise.

Build optionality into your products, into your pricing, into your org design, and then pay attention to the leading indicators across all of them. The old five year plan, refreshed every year, dead. So I believe AI could lead to a software renaissance. There's so much opportunity out there.

Probabilistic and deterministic hybrids, hyper-niche software that suddenly can serve, you know, 10 customers, 100 customers in a profitable business. Malleable and self-evolving software, we haven't invented the new UX yet. Of course, we'll be building agents, but we'll also be building software for agents. All of this software.

All of this ⁓ is for us in this room to build. But to get there we have to think differently, we have to think bigger, and we have to bring our organisations along with us. So Geoffrey Hinton gave radiologists five years and instead they got a renaissance. So my challenge to you as you leave the conference today is what are you going to do with yours?

Obrigado Lisboa.